Dynamic simulation is an accurate but time-consuming way tu calculate energy and thermal comfort performances of passive solar buildings. Simplified methods, originally developed to evercome their disadvantage, unfortunately do not provide the same quantitative and qualitative results, particulary concerning thermal comfort. A new approach to thermal building simulation, based on Markov chains, has been proposed which theoretically combines the advantages of the two former computing techniques. Succesful validation of the Markovian stochastic model has been carried out on different passive solar devices. A main difficulty of the methos is the digitizing of the ddriving and internal variables which is necessary to set up the Markov model. Systematic procedures have been developed to carry out the digitizing in order to avoid adding supplementary difficulties to the simulation process when compared to conventional methods. However, the procedure leads to an important increase of computer memory requirement. Possibilities and limitations of the Markovian stochastic model will be discussed in this paper . Procedures used in order to facilitate the use of the method in practice will be presented, in particular with regard to implementation on conventionel PC's.